Real-time AI matte enhancement · Live virtual production

NullScreen AI

The compositor sees the same streams it always did — the matte just got better.

A decoupled, real-time AI refinement node for live broadcast virtual production. It drops into the path between your hardware keyer and compositor, sharpens the matte on GPU, and hands it back. No workflow change for the operator.

Get in touch In development · 2026

The node

Four things a hardware keyer alone can't do

Added in real time, on GPU, inside a live broadcast frame budget — starting from the clean matte your keyer already produces.

Temporal

Motion-adaptive matte

Per-region temporal blending kills flicker and ghosting at once — no studio-wide flicker-versus-smear dial to accept.

Spill

AI despill

Color-accurate removal of green and blue spill that preserves true skin and material tone.

Plate

Per-frame clean plate

Estimates the background behind talent for better spill correction, relighting headroom, and fallback keying.

Shadow

Contact-shadow synthesis

Camera-tracking-driven soft shadow where talent meets the virtual floor — something an image-only matte can't produce.

Why it's different

Decoupled. Augmenting. Adaptive.

Decoupled

Not captive

Works with the keyer and compositor you already own. No proprietary engine, no same-box requirement, no rip-and-replace.

Augment

Refine, don't replace

We refine the clean matte your hardware keyer already produces, and let AI attack only the weak parts — spill, plate, shadow, motion.

Adaptive

Per-region temporal

Memory driven per-pixel by motion, instead of one global dial that forces a flicker-versus-smear choice on the entire studio.

Roadmap

From augment to markerless

Land studios on the low-risk augment step today, then carry more of the keying load over time — with motion-adaptive temporal as the backbone the whole way.

01Now

Augment

Refine the matte your hardware keyer already produces. Zero workflow change, immediate quality lift.

02Next

Reduce green-screen

AI carries more of the load, hybridizing with whatever key and tracking signal the room already has.

03Later

Markerless

Full AI matting with motion-adaptive temporal as the backbone. No green screen at all.

Founder

Ulf Soderqvist

Director of Photography, Founder and Application Architect

Builds the entire stack — concept, code, and camera. Based in Washington State (USA) and Sweden.

Contact

Let's talk.

For demos, partnerships, or early access to the refinement node.

ulf@nullscreen.ai

HoodAngelZ LLC
1309 Coffeen Avenue, STE 1200
Sheridan, WY 82801, USA